Search

Probabilistic inductive logic programming (PILP) is a statistical relational learning technique which extends inductive logic programming by considering probabilistic data. The ability to use probabilities to represent uncertainty comes at the cost of an exponential evaluation time when composing theories to model the given problem. For this reason, PILP systems rely on various pruning strategies in order to reduce the search space. However, to the best of the authors’ knowledge, there has been no systematic analysis of the different pruning strategies, how they impact the search space and how they interact with one another. This work presents a unified representation for PILP pruning strategies which enables end-users to understand how these strategies work both individually and combined and to make an informed decision on which pruning strategies to select so as to best achieve their goals. The performance of pruning strategies is evaluated both time and quality-wise in two state-of-the-art PILP systems with datasets from three different domains. Besides analysing the performance of the pruning strategies, we also illustrate the utility of PILP in one of the application domains, which is a real-world application.

The G-Skyline (GSky) query is formulated to report optimal groups that are not dominated by any other group of the same size. Particularly, a given group $$G_1$$ dominates another group $$G_2$$ if for any point $$p\in G_1$$ , p dominates or equals to points $$p{'}\in G_2$$ ; at the same time, there is at least one point p dominating $$p{'}$$ . Most existing group skyline queries need to calculate an aggregate point for each group. Compared to these queries, the GSky query is more practical because it avoids specifying an aggregate function which leads to miss important results containing non-skyline points. This means the GSky query can get much more comprehensive query results which not only contain the G-Skylines consisting of skyline points but also the G-Skylines including non-skyline points. Here, a non-skyline point is dominated by another point in a given data set. However, the GSky query usually returns too many results, making it a big burden for users to pick out their expected results. To address these issues, we investigate a flexible group skyline query, namely Flexible G-Skyline (FGSky) query, which is flexible and practical for directly computing the optimal groups on the basis of user preferences. In this paper, we formulate the FGSky query, identify its properties, and present effective pruning strategies. Besides, we propose progressive algorithms for the FGSky query where a grouping strategy and a layered strategy are utilized to get better query performance. Through extensive experiments on both synthetic and real data sets, we demonstrate the efficiency, effectiveness, and progressiveness of the proposed algorithms.

International audience; This paper presents parallel solutions (developed based on two state-of-the-art algorithms iSAX and sketch) for evaluating kNN (k nearest neighbor) queries on large databases of time series, compares them based on various measures of quality and time performance, offers a tool that uses the characteristics of application data to determine which algorithm to choose for that application and how to set the parameters for that algorithm. Specifically, our experiments show that: (i) iSAX and its derivatives perform best in both time and quality when the time series can be characterized by a few low frequency Fourier Coefficients , a regime where the iSAX pruning approach works well. (ii) iSAX performs significantly less well when high frequency Fourier Coefficients have much of the energy of the time series. (iii) A random projection approach based on sketches by contrast is more or less independent of the frequency power spectrum. The experiments show the close relationship between pruning ratio and time for exact iSAX as well as between pruning ratio and the quality of approximate iSAX. Our toolkit analyzes typical time series of an application (i) to determine optimal segment sizes for iSAX and (ii) when to use Parallel Sketches instead of iSAX. Our algorithms have been implemented using Spark, evaluated over a cluster of nodes, and have been applied to both real and synthetic data. The results apply to any databases of numerical sequences, whether or not they relate to time.

High utility itemset mining is an important extension of frequent itemset mining which considers unit profits and quantities of items as external and internal utilities, respectively. Since the utility function has not downward closure property, an overestimated value of utility is obtained using an anti-monotonic upper bound of utility function to prune the search space and improve the efficiency of high utility itemset mining methods. Transaction-weighted utilization (TWU) of itemset was the first and one of the most important functions which has been used as the anti-monotonic upper bound of utility by various algorithms. A variety of high utility itemset mining methods have attempted to tighten the utility upper bound and have exploited appropriate pruning strategies to improve mining efficiency. Although TWU and its improved alternatives have attempted to increase the efficiency of high utility itemset mining methods by pruning their search spaces, they suffer from a significant number of generated candidates which are high-TWU but are not high utility itemsets. Calculating the actual utilities of low utility candidates needs to multiple scanning of the dataset and thus imposes a huge overhead to the mining methods, which can cause to lose the pruning benefits of the upper bounds. Proposing appropriate pruning strategies, exploiting efficient data structures, and using tight anti-monotonic upper bounds can overcome this problem and lead to significant performance improvement in high utility itemset mining methods. In this paper, a new projection-based method, called MAHI (matrix-aided high utility itemset mining), is introduced which uses a novel utility matrix-based pruning strategy, called MA-prune to improve the high utility itemset mining performance in terms of execution time. The experimental results show that MAHI is faster than former algorithms.

The emergence and proliferation of the internet of things (IoT) devices have resulted in the generation of big and uncertain data due to the varied accuracy and decay of sensors and their different sensitivity ranges. Since data uncertainty plays an important role in IoT data, mining the useful information from uncertain dataset has become an important issue in recent decades. Past works focus on mining the high sequential patterns from the uncertain database. However, the utility of a derived sequence increases along with the size of the sequence, which is an unfair measure to evaluate the utility of a sequence since any combination of a high-utility sequence will also be the high-utility sequence, even though the utility of a sequence is merely low. In this paper, we address the limitation of the previous potential high-utility sequential pattern mining and present a potentially high average-utility sequential pattern mining framework for discovering the set of potentially high average-utility sequential patterns (PHAUSPs) from the uncertain dataset by considering the size of a sequence, which can provide a fair measure of the patterns than the previous works. First, a baseline potentially high average-utility sequential pattern algorithm and three pruning strategies are introduced to completely mine the set of the desired PHAUSPs. To reduce the computational cost and accelerate the mining process, a projection algorithm called PHAUP is then designed, which leads to a reduction in the size of candidates of the desired patterns. Several experiments in terms of runtime, number of candidates, memory overhead, number of discovered pattern, and scalability are then evaluated on both real-life and artificial datasets, and the results showed that the proposed algorithm achieves promising performance, especially the PHAUP approach.

The problem of identifying duplicated entities in a dataset has gained increasing importance during the last decades. Due to the large size of the datasets, this problem can be very costly to be solved due to its intrinsic quadratic complexity. Both researchers and practitioners have developed a variety of techniques aiming to speed up a solution to this problem. One of these techniques is called blocking, an indexing technique that splits the dataset into a set of blocks, such that each block contains entities that share a common property evaluated by a blocking key function. In order to improve the efficacy of the blocking technique, multiple blocking keys may be used, and thus, a set of blocking results is generated. In this paper, we investigate how to control the size of the blocks generated by the use of multiple blocking keys and maintain reasonable quality results, which is measured by the quality of the produced blocks. By controlling the size of the blocks, we can reduce the overall cost of solving an entity resolution problem and facilitate the execution of a variety of tasks (e.g., real-time and privacy-preserving entity resolution). For doing so, we propose many heuristics which exploit the co-occurrence of entities among the generated blocks for pruning, splitting and merging blocks. The experimental results we carry out using four datasets confirm the adequacy of the proposed heuristics for generating block sizes within a predefined range threshold as well as maintaining reasonable blocking quality results.

Similarity join on time series databases is an essential operation for data analysis applications. Due to the curse of dimensionality, it is not suitable to use traditional index techniques, such as R-tree and kd-tree. In the paper, a dynamic segment index (i.e., DSTree) is utilized to reduce the huge comparison cost on the similarity join on time series databases. However, the DSTree is designed for similarity search and only supports bound estimations between a time series and a batch of time series in a DSTree node. To make the DSTree suitable for the similarity join on time series databases, it is necessary to have tight bounds for the nodes to achieve a better pruning power, where the biggest challenge is that the DSTree nodes may have different segmentations. To solve the problem aforementioned, a segmentation alignment and synopsis evaluation method is proposed to support the estimation of DSTree nodes to significantly reduce the time cost by pruning unnecessary comparisons. Moreover, to make our approach I/O efficient, a caching strategy is proposed by taking advantage of both graph partitioning and the locality of the DSTree index. The efficiency and effectiveness of the proposed approaches are verified by experiments on real-life datasets.

Efficient and accurate near-duplicate detection is a trending topic of research. Complications arise from the great time and space complexities of existing algorithms. This study proposes a novel pruning strategy to improve pairwise comparison-based near-duplicate detection methods. After parsing the documents into punctuation-delimited blocks called chunks, it decides between the categories of “near duplicate,” “non-duplicate” or “suspicious” by applying certain filtering rules. This early decision makes it possible to disregard many of the non-necessary computations—on average 92.95% of them. Then, for the suspicious pairs, common chunks and short chunks are removed and the remaining subsets are reserved for near-duplicate detection. Size of the remaining subsets is on average 4.42% of the original corpus size. Evaluation results show that near-duplicate detection with the proposed strategy in its best configuration (CHT = 8, τ = 0.1) has F-measure = 87.22% (precision = 86.91% and recall = 87.54%). Its F-measure is comparable with the SpotSig method with less execution time. In addition, applying the proposed strategy in a near-duplicate detection process eliminates the need for preprocessing. It is also tunable to achieve the intended levels of near duplication and noise suppression.

In this paper, we present a novel fuzzy rule extraction approach by employing the Gaussian kernels and fuzzy concept lattices. First we introduce the Gaussian kernel to interval type-2 fuzzy rough sets to model fuzzy similarity relations and introduce a few concepts and theorems to improve the classification performance with fewer attributes accordingly. Based on this idea, we propose a novel attribute reduction algorithm, which can achieve better classification performance of deducing reduction subset of fewer attributes, and this will be used in the subsequent decision rule extraction. Then we justify the necessary and sufficient conditions of our fuzzy rule extraction approach through three implicit rule theorems and present a novel fuzzy decision rule extraction algorithm using fuzzy concept lattices and introduce the concepts of frequent nodes and candidate 2-tuples to our pruning strategy. Also, comparative performance experiments are carried out on the UCI datasets, and the results of both reduction subset size and classification ability show the advantages of our algorithm.

© 2019, Springer-Verlag London Ltd., part of Springer Nature. The popularity of e-business and preference learning techniques have contributed a huge amount of product and user preference data. Analyzing the influence of an existing or new product among the users is critical to unlock the great scientific and social–economic value of these data. In this paper, we advocate the problem of influence-based cost optimization for the user preference and product data, which is fundamental in many real applications such as marketing and advertising. Generally, we aim to find a cost optimal position for a new product such that it can attract at least k or a particular percentage of users for the given user preference functions and competitors’ products. Although we show the solution space of our problem can be reduced to a finite number of possible positions (points) by utilizing the classical k-level computation techniques, the computation cost is still very expensive due to the nature of the high combinatorial complexity of the k-level problem. To alleviate this issue, we develop efficient pruning and query processing techniques to significantly improve the performance. In particular, our traverse-based 2-dimensional algorithm is very efficient with time complexity O(n) where n is the number of user preference functions. For general multi-dimensional spaces, we develop space partition-based algorithm to significantly improve the performance by utilizing cost-based, influence-based and local dominance-based pruning techniques. Then, we show that the performance of the partition-based algorithm can be further enhanced by utilizing sampling approach, where the problem can be reduced to the classical half-space intersection problem. Based on the problem of influence-based cost optimization, two naturally extended problems are proposed, which are Batch-Query and the most cost-effective influence query. We demonstrate the efficiency of our techniques with extensive experiments over real and synthetic datasets.

Named entity recognition is a vital task for various applications related to biomedical natural language processing. It aims at extracting different biomedical entities from the text and classifying them into some predefined categories. The types could vary depending upon the genre and domain, such as gene versus non-gene in a coarse-grained scenario, or protein, DNA, RNA, cell line, and cell-type in a fine-grained scenario. In this paper, we present a novel filter-based feature selection technique utilizing the search capability of particle swarm optimization (PSO) for determining the most optimal feature combination. The technique yields in the most optimized feature set, that when used for classifiers learning, enhance the system performance. The proposed approach is assessed over four popular biomedical corpora, namely GENIA, GENETAG, AIMed, and Biocreative-II Gene Mention Recognition (BC-II). Our proposed model obtains the F score values of $$74.49\%$$ , $$91.11\%$$ , $$90.47\%$$ , $$88.64\%$$ on GENIA, GENETAG, AIMed, and BC-II dataset, respectively. The efficiency of feature pruning through PSO is evident with significant performance gains, even with a much reduced set of features.

In many applications, top-k skyline query is an important operation to return k skyline tuples with the highest domination scores in a potentially huge data space. It is analyzed that the existing algorithms cannot process top-k skyline query on massive data efficiently. In this paper, we propose a novel table-scan-based algorithm RSTS to compute top-k skyline results on massive data efficiently. RSTS first builds the presorted table, whose tuples are arranged in the order of round-robin retrieval on sorted column lists. RSTS consists of two phases. In phase 1, the candidate tuples are acquired by the sequential scan on the presorted table. In phase 2, RSTS calculates the domination scores of the candidates and returns query results by another sequential scan. It is proved that RSTS has the characteristic of early termination, along with the theoretical analysis of scan depths. The pruning rule for candidate tuples is devised in this paper. The theoretical pruning effect shows that majority of the skyline results can be discarded directly. The extensive experimental results, conducted on synthetic and real-life data sets, show that RSTS outperforms the existing algorithms significantly.

In many applications, dynamic skyline query is an important operation to find the interesting tuples in a potentially huge data space. Given the query point, dynamic skyline query returns tuples which are not dynamically dominated by other tuples. It is found that the existing algorithms cannot process dynamic skyline query on massive data efficiently. This paper proposes a novel dynamic-sorted-list-based DDS algorithm to efficiently compute dynamic skyline results on massive data. Given the query point, the dynamic sorted list of each attribute is not materialized but generated dynamically by the sorted list of the attribute. DDS retrieves the tuples in the involved dynamic sorted lists in the round-robin fashion until the early termination condition is satisfied, and computes the dynamic skyline results by retrieving the candidates. The pruning operation is devised to reduce the number of the retrieved candidates. The extensive experimental results, conducted on synthetic and real-life data sets, show that DDS outperforms the existing algorithms significantly.

Dynamic Time Warping (DTW) is a popular method for measuring the similarity of time series. It is widely used in various domains. A major drawback of DTW is that it has a high computational complexity. To address this problem, pruning techniques to calculate the exact DTW distance, as well as DTW approximation methods, have become important approaches. In this paper, we introduce Blocked Dynamic Time Warping (BDTW), a new similarity measure which works on run-length encoded time series representation. BDTW utilizes any repetitive values (zero and nonzero) in time series to reduce DTW computation time. BDTW closely approximates DTW distance, and it is significantly faster than traditional DTW for time series with high levels of value repetition. Moreover, BDTW can be combined with time series representation methods which provide constant segments, to serve as a close approximation method even for the time series without value repetition. Constrained BDTW, BDTW upper bound and BDTW lower bound are discussed as variations of BDTW. BDTW upper bound and BDTW lower bound are presented as a new DTW upper bound and lower bound which can be efficiently applied on time series with high levels of value repetition for pruning unhopeful alignments and matches in the exact DTW calculation. We show the effectiveness of BDTW and its variations on different applications using the following datasets: Almanac of Minutely Power, Refit Smart Homes, as well as the 85 datasets from the University of California, Riverside time series classification archive (UCR archive).

Mining frequent sequences in sequential databases are highly valuable for many real-life applications. However, in several cases, especially when databases are huge and when low minimum support thresholds are used, the cardinality of the result set can be enormous. Consequently, algorithms for discovering frequent sequences exhibit poor performance, showing an important increase in execution time, memory consumption and storage space usage. To address this issue, researchers have studied the tasks of mining frequent closed and generator sequences, as they provide several benefits when compared to the set of frequent sequences. One of the most important benefits is that the cardinalities of frequent closed and generator sequences are generally much less than the cardinality of frequent sequences. Hence, humans find it more convenient to analyze the information provided by closed and generator sequences. Moreover, it was shown that frequent closed sequences have the advantage of being lossless, and they thus preserve information about the frequency of all frequent subsequences, while generator sequences can provide higher accuracy for sequence classification tasks since they are the smallest patterns that characterize groups of sequences. Besides, frequent closed sequences can be combined with generators to produce non-redundant sequential rules and recover the complete set of frequent sequences and their frequencies. This paper proposes two novel algorithms named FCloSM and FGenSM to mine frequent closed and generator sequences efficiently. These algorithms are based on new pruning conditions called extended early elimination (3E) and early pruning techniques named EPCLO and EPGEN, designed to identify non-closed and non-generator patterns early. Based on these techniques, two local pruning strategies called LPCLO and LPGEN are proposed to eliminate non-closed and non-generator patterns more efficiently at two successive levels of the prefix search tree without performing subsequence relation checking. These theoretical results, which are the basis of FCloSM and FGenSM, are mathematically proved and are shown to be more general than those presented in previous work. Extensive experiments show that FCloSM and FGenSM are one to two orders of magnitude faster than the state-of-the-art algorithms for discovering frequent closed sequences (CloSpan, BIDE, ClaSP and CM-ClaSP) and for mining frequent generators (FEAT, FSGP and VGEN), and that FCloSM and FGenSM consume much less memory.

Recently, high utility pattern mining (HUPM) has been extensively studied. Many approaches for HUPM have been proposed in recent years, but most of them aim at mining HUPs without any consideration for their frequency. This has the major drawback that any combination of a low utility item with a very high utility pattern is regarded as a HUP, even if this combination has low affinity and contains items that rarely co-occur. Thus, frequency should be a key criterion to select HUPs. To address this issue, and derive high utility interesting patterns (HUIPs) with strong frequency affinity, the HUIPM algorithm was proposed. However, it recursively constructs a series of conditional trees to produce candidates and then derive the HUIPs. This procedure is time-consuming and may lead to a combinatorial explosion when the minimum utility threshold is set relatively low. In this paper, an efficient algorithm named fast algorithm for mining discriminative high utility patterns (DHUPs) with strong frequency affinity (FDHUP) is proposed to efficiently discover DHUPs by considering both the utility and frequency affinity constraints. Two compact structures named EI-table and FU-tree and three pruning strategies are introduced in the proposed algorithm to reduce the search space, and efficiently and effectively discover DHUPs. An extensive experimental study shows that the proposed FDHUP algorithm considerably outperforms the state-of-the-art HUIPM algorithm in terms of execution time, memory consumption, and scalability.

We present Silverback+, a scalable probabilistic framework for accurate association rule and frequent item-set mining of large-scale social behavioral data. Silverback+ tackles the problem of efficient storage utilization and management via: (1) probabilistic columnar infrastructure and (2) using Bloom filters and sampling techniques. In addition, probabilistic pruning techniques based on Apriori method are developed, for accelerating the mining of frequent item-sets. The proposed target-driven techniques yield a significant reduction of the size of the frequent item-set candidates, as well as the required number of repetitive membership checks through a novel list intersection algorithm. Extensive experimental evaluations demonstrate the benefits of this context-aware consideration and incorporation of the infrastructure limitations when utilizing the corresponding research techniques. When compared to the traditional Hadoop-based approach for improving scalability by straightforwardly adding more hosts, Silverback+ exhibits a much better runtime performance, with negligible loss of accuracy.

High utility sequential pattern mining is an emerging topic in pattern mining, which refers to identify sequences with high utilities (e.g., profits) but probably with low frequencies. To identify high utility sequential patterns, due to lack of downward closure property in this problem, most existing algorithms first generate candidate sequences with high sequence-weighted utilities (SWUs), which is an upper bound of the utilities of a sequence and all its supersequences, and then calculate the actual utilities of these candidates. This causes a large number of candidates since SWU is usually much larger than the real utilities of a sequence and all its supersequences. In view of this, we propose two tight utility upper bounds, prefix extension utility and reduced sequence utility, as well as two companion pruning strategies, and devise HUS-Span algorithm to identify high utility sequential patterns by employing these two pruning strategies. In addition, since setting a proper utility threshold is usually difficult for users, we also propose algorithm TKHUS-Span to identify top-k high utility sequential patterns by using these two pruning strategies. Three searching strategies, guided depth-first search (GDFS), best-first search (BFS) and hybrid search of BFS and GDFS, are also proposed to improve the efficiency of TKHUS-Span. Experimental results on some real and synthetic datasets show that HUS-Span and TKHUS-Span with strategy BFS are able to generate less candidate sequences and thus outperform other prior algorithms in terms of mining efficiency.

Time-series classification is an important problem for the data mining community due to the wide range of application domains involving time-series data. A recent paradigm, called shapelets, represents patterns that are highly predictive for the target variable. Shapelets are discovered by measuring the prediction accuracy of a set of potential (shapelet) candidates. The candidates typically consist of all the segments of a dataset; therefore, the discovery of shapelets is computationally expensive. This paper proposes a novel method that avoids measuring the prediction accuracy of similar candidates in Euclidean distance space, through an online clustering/pruning technique. In addition, our algorithm incorporates a supervised shapelet selection that filters out only those candidates that improve classification accuracy. Empirical evidence on 45 univariate datasets from the UCR collection demonstrates that our method is 3---4 orders of magnitudes faster than the fastest existing shapelet discovery method, while providing better prediction accuracy. In addition, we extended our method to multivariate time-series data. Runtime results over four real-life multivariate datasets indicate that our method can classify MB-scale data in a matter of seconds and GB-scale data in a matter of minutes. The achievements do not compromise quality; on the contrary, our method is even superior to the multivariate baseline in terms of classification accuracy.

In many applications, top-k query is an important operation to return a set of interesting points in a potentially huge data space. The existing algorithms, either maintaining too many candidates, or requiring assistant structures built on the specific attribute subset, or returning results with probabilistic guarantee, cannot process top-k query on massive data efficiently. This paper proposes a sorted-list-based TKAP algorithm, which utilizes some data structures of low space overhead, to efficiently compute top-k results on massive data. In round-robin retrieval on sorted lists, TKAP performs adaptive pruning operation and maintains the required candidates until the stop condition is satisfied. The adaptive pruning operation can be adjusted by the information obtained in round-robin retrieval to achieve a better pruning effect. The adaptive pruning rule is developed in this paper, along with its theoretical analysis. The extensive experimental results, conducted on synthetic and real-life data sets, show the significant advantage of TKAP over the existing algorithms.

In this paper, we propose a new type of queries to retrieve the top-k most influential locations from a candidate set $$C$$ C given sets of customers $$M$$ M and existing facilities $$F$$ F . The influence models the popularity of a facility. Such queries have wide applications in decision support systems. A naive solution sequentially scans (SS) all data sets, which is expensive, and hence, we investigate two branch-and-bound algorithms for the query, namely Estimate Expanding Pruning (EEP) and Bounding Influence Pruning (BIP). Both algorithms follow the best first traverse. On determining the traversal order, while EEP leverages distance metrics between nodes, BIP relies on half plane pruning which avoids the repetitive estimations in EEP. As our experiments shown, BIP is much faster than SS which outperforms EEP, while the worst-case complexity of EEP and BIP is worse than that of SS. To improve the efficiency, we further propose a Nearest Facility Circle Join (NFCJ) algorithm. NFCJ builds an influence R-tree on the influence relationship between customers and existing facilities and joins the candidate R-tree with the influence R-tree to obtain the results. We compare all algorithms and conclude that NFCJ is the best solution, which outperforms SS, EEP, and BIP by orders of magnitude.

In many applications including multicriteria decision making, top-k dominating query is a practically useful tool to return k tuples with the highest domination scores in a potentially huge data space. The existing algorithms, either requiring indexes built on the specific attribute subset, or incurring high I/O cost or memory cost, cannot process top-k dominating query on massive data efficiently. In this paper, a novel algorithm TDEP is proposed to utilize sorted lists built for each attribute with low cost to return top-k dominating result on massive data efficiently. Through analysis, it is found that TDEP can be divided into two phases: growing phase and shrinking phase. In each phase, TDEP retrieves the sorted lists in round-robin fashion and maintains the candidates until the stop condition is satisfied. The theoretical analysis is provided for the execution behavior in two phases. An efficient method is developed to compute the domination scores of tuples with the obtained candidates only. Besides, TDEP adopts early pruning to reduce the number of candidate tuples maintained significantly. The extensive experimental results, conducted on synthetic and real-life data sets, show the significant performance advantage of TDEP over the existing algorithms.

Finding an optimal meeting point (OMP) for a group of people (or a set of objects) at different locations is an important problem in spatial query processing. There are many real-life applications related to this problem, such as determining the location of a conference venue, deciding the pick-up location of a tourist bus, and planing tactics of artificial intelligence in real-time strategy games. Formally, given a set $$Q$$ Q of query points in a spatial setting $$P$$ P , an OMP query fetches the point $$o\in P$$ o ? P that minimizes a cost function defined over the distances from $$o$$ o to all points in $$Q$$ Q . Since there are infinitely many locations in a given space setting, it is infeasible to examine all of them to find the OMP, and thus, the problem is challenging. In this paper, we study OMP queries in the following two spatial settings which are common in real-life applications: Euclidean space and road networks. In the setting of Euclidean space, we propose a general framework for answering all OMP query variants and also identify the best algorithms for particular types of OMP queries in the literature. In the setting of road networks, we study how to access only part of the road network and examine part of the candidates. Specifically, we explore two pruning techniques, namely Euclidean distance bound and threshold algorithm, which help improve the efficiency of OMP query processing. Extensive experiments are conducted to demonstrate the efficiency of our proposed approaches on both real and synthetic datasets.

Recently, utility mining has widely been discussed in the field of data mining. It finds high utility itemsets by considering both profits and quantities of items in transactional data sets. However, most of the existing approaches are based on the principle of levelwise processing, as in the traditional two-phase utility mining algorithm to find a high utility itemsets. In this paper, we propose an efficient utility mining approach that adopts an indexing mechanism to speed up the execution and reduce the memory requirement in the mining process. The indexing mechanism can imitate the traditional projection algorithms to achieve the aim of projecting sub-databases for mining. In addition, a pruning strategy is also applied to reduce the number of unpromising itemsets in mining. Finally, the experimental results on synthetic data sets and on a real data set show the superior performance of the proposed approach.

A hyperclique (Xiong et al. in Proceedings of the IEEE international conference on data mining, pp 387---394, 2003) is an itemset containing items that are strongly correlated with each other, based on a user-specified threshold. Hypercliques (HCs) have been successfully used in a number of applications, for example, clustering (Xiong et al. in Proceedings of the 4th SIAM international conference on data mining, pp 279---290, 2004) and noise removal (Xiong et al. in IEEE Trans Knowl Data Eng 18(3):304---319, 2006). Even though HC has been shown to respond well to datasets with skewed support distribution and low support threshold, it may still grow very large for dense datasets and lower h-confidence threshold. In this paper, we propose a new pruning method based on combining HCs and non-derivable itemsets (NDIs) (Calders and Goethals in Proceedings of the PKDD international conference on principles of data mining and knowledge discovery, pp 74---85, 2002) in order to substantially reduce the amount of generated HCs. Specifically, we propose a new collection of HCs, called non-derivable hypercliques (NDHCs). The NDHC collection is a lossless representation of HCs, that is, given the itemsets in NDHCs, we can generate the complete HC collection and their support, without additional scanning of the dataset. We present an efficient algorithm to mine all NDHC sets, NDHCMiner, and an algorithm to derive all HC sets and their support from NDHCs, NDHCDeriveAll. We experimentally compare our collection, NDHC with HC, with respect to runtime performance as well as total number of generated sets, using real and artificial data. We also show comparisons with another condensed representation of HCs, maximal hyperclique patterns (MHPs). Our experiments show that the NDHC collection offers substantial advantages over HCs, and even MHPs, especially for dense datasets and lower h-confidence values.

We introduce the problem of diverse dimension decomposition in transactional databases, where a dimension is a set of mutually exclusive itemsets. The problem we consider requires to find a decomposition of the itemset space into dimensions, which are orthogonal to each other and which provide high coverage of the input database. The mining framework we propose can be interpreted as a dimensionality-reducing transformation from the space of all items to the space of orthogonal dimensions. Relying on information-theoretic concepts, we formulate the diverse dimension decomposition problem with a single objective function that simultaneously captures constraints on coverage, exclusivity, and orthogonality. We show that our problem is NP-hard, and we propose a greedy algorithm exploiting the well-known FP-tree data structure. Our algorithm is equipped with strategies for pruning the search space deriving directly from the objective function. We also prove a property that allows assessing the level of informativeness for newly added dimensions, thus allowing to define criteria for terminating the decomposition. We demonstrate the effectiveness of our solution by experimental evaluation on synthetic datasets with known dimension and three real-world datasets, flickr, del.icio.us and dblp. The problem we study is largely motivated by applications in the domain of collaborative tagging; however, the mining task we introduce in this paper is useful in other application domains as well.

Reverse nearest-neighbor (RNN) query processing is important for many applications such as decision-support systems, profile-based marketing and molecular biology; consequently, RNN query processing has attracted considerable attention in the research community in recent years. Most existing approaches for RNN query processing either rely on nearest-neighbor pre-computation or work for specific data space (e.g., the Euclidean space). The only method for RNN query processing in metric space is based on the M-tree. In this paper, we propose an approach for RNN query processing in high-dimensional metric space using distance-based index structure (in particular, NAQ-tree that outperforms the other distance-based index structures as we have already verified in a previous study). In high-dimensional space, the properties of distance-based index structure provide strong pruning rules than the M-tree. In addition, unlike the previous work, our approach integrates the filtering and verification steps and uses the information obtained in the verification stage to further improve the filtering rate. Our approach delivers results incrementally and hence well serves real-time applications. The reported experimental results demonstrate the applicability and effectiveness of the proposed NAQ-tree-based RNN approach.

Most of the current algorithms for mining frequent patterns assume that two object subdescriptions are similar if they are equal, but in many real-world problems some other ways to evaluate the similarity are used. Recently, three algorithms (ObjectMiner, STreeDC-Miner and STreeNDC-Miner) for mining frequent patterns allowing similarity functions different from the equality have been proposed. For searching frequent patterns, ObjectMiner and STreeDC-Miner use a pruning property called Downward Closure property, which should be held by the similarity function. For similarity functions that do not meet this property, the STreeNDC-Miner algorithm was proposed. However, for searching frequent patterns, this algorithm explores all subsets of features, which could be very expensive. In this work, we propose a frequent similar pattern mining algorithm for similarity functions that do not meet the Downward Closure property, which is faster than STreeNDC-Miner and loses fewer frequent similar patterns than ObjectMiner and STreeDC-Miner. Also we show the quality of the set of frequent similar patterns computed by our algorithm with respect to the quality of the set of frequent similar patterns computed by the other algorithms, in a supervised classification context.

Frequent itemset mining over data streams becomes a hot topic in data mining and knowledge discovery in recent years, and has been applied to different areas. However, the setting of a minimum support threshold needs some domain knowledge. Itwill bring a lot of difficulties or much burden to users if the support threshold is not set reasonably. It is interesting for users to find top-K frequent itemsets over data streams. In this paper, a dynamical incremental approximate algorithm TOPSIL-Miner is presented to mine top-K significant itemsets in landmark windows. A new data structure, TOPSIL-Tree, is designed to store the potential significant itemsets and other data structures of maximum support list, ordered item list, TOPSET and minimum support list are devised to maintain information about mining results. Moreover, three optimal strategies are exploited to reduce time and space cost of the algorithm: (1) pruning trivial nodes in the current data stream, (2) promoting ining support threshold during mining process adaptively and heuristically, and (3) promoting pruning threshold dynamically. The accuracy of the algorithm is also analyzed. Extensive experiments are performed to evaluate the good effectiveness and the high efficiency and precision of the algorithm.

The problem of discovering frequent arrangements of temporal intervals is studied. It is assumed that the database consists of sequences of events, where an event occurs during a time-interval. The goal is to mine temporal arrangements of event intervals that appear frequently in the database. The motivation of this work is the observation that in practice most events are not instantaneous but occur over a period of time and different events may occur concurrently. Thus, there are many practical applications that require mining such temporal correlations between intervals including the linguistic analysis of annotated data from American Sign Language as well as network and biological data. Three efficient methods to find frequent arrangements of temporal intervals are described; the first two are tree-based and use breadth and depth first search to mine the set of frequent arrangements, whereas the third one is prefix-based. The above methods apply efficient pruning techniques that include a set of constraints that add user-controlled focus into the mining process. Moreover, based on the extracted patterns a standard method for mining association rules is employed that applies different interestingness measures to evaluate the significance of the discovered patterns and rules. The performance of the proposed algorithms is evaluated and compared with other approaches on real (American Sign Language annotations and network data) and large synthetic datasets.

Given a set of k-dimensional objects, the skyline query finds the objects that are not dominated by others. In practice, different users may be interested in different dimensions of the data, and issue queries on any subset of k dimensions in stream environments. This paper focuses on supporting concurrent and unpredictable subspace skyline queries over data streams. Simply to compute and store the skyline objects of every subspace in stream environments will incur expensive update cost. To balance the query cost and update cost, we only maintain the full space skyline in this paper. We first propose an efficient maintenance algorithm and several novel pruning techniques. Then, an efficient and scalable two-phase algorithm is proposed to process the skyline queries in different subspaces based on the full space skyline. Furthermore, we present the theoretical analyses and extensive experiments that demonstrate our method is both efficient and effective.

Detecting association rules with low support but high confidence is a difficult data mining problem. To find such rules using approaches like the Apriori algorithm, minimum support must be set very low, which results in a large number of redundant rules. We are interested in sporadic rules; i.e. those that fall below a maximum support level but above the level of support expected from random coincidence. There are two types of sporadic rules: perfectly sporadic and imperfectly sporadic. Here we are more concerned about finding imperfectly sporadic rules, where the support of the antecedent as a whole falls below maximum support, but where items may have quite high support individually. In this paper, we introduce an algorithm called Mining Interesting Imperfectly Sporadic Rules (MIISR) to find imperfectly sporadic rules efficiently, e.g. fever, headache, stiff neck → meningitis. Our proposed method uses item constraints and coincidence pruning to discover these rules in reasonable time. This paper is an expanded version of Koh et al. [Advances in knowledge discovery and data mining: 10th Pacific-Asia Conference (PAKDD 2006), Singapore. Lecture Notes in Computer Science 3918, Springer, Berlin, pp 473–482].

In this paper, we propose an efficient scalable algorithm for mining Maximal Sequential Patterns using Sampling (MSPS). The MSPS algorithm reduces much more search space than other algorithms because both the subsequence infrequency-based pruning and the supersequence frequency-based pruning are applied. In MSPS, a sampling technique is used to identify long frequent sequences earlier, instead of enumerating all their subsequences. We propose how to adjust the user-specified minimum support level for mining a sample of the database to achieve better overall performance. This method makes sampling more efficient when the minimum support is small. A signature-based method and a hash-based method are developed for the subsequence infrequency-based pruning when the seed set of frequent sequences for the candidate generation is too big to be loaded into memory. A prefix tree structure is developed to count the candidate sequences of different sizes during the database scanning, and it also facilitates the customer sequence trimming. Our experiments showed MSPS has very good performance and better scalability than other algorithms.

Discovering contrasts between collections of data is an important task in data mining. In this paper, we introduce a new type of contrast pattern, called a minimal distinguishing subsequence (MDS). An MDS is a minimal subsequence that occurs frequently in one class of sequences and infrequently in sequences of another class. It is a natural way of representing strong and succinct contrast information between two sequential datasets and can be useful in applications such as protein comparison, document comparison and building sequential classification models. Mining MDS patterns is a challenging task and is significantly different from mining contrasts between relational/transactional data. One particularly important type of constraint that can be integrated into the mining process is the maximum gap constraint. We present an efficient algorithm called ConSGapMiner, to mine all MDSs according to a maximum gap constraint. It employs highly efficient bitset and Boolean operations, for powerful gap based pruning within a prefix growth framework. A performance evaluation with both sparse and dense datasets, demonstrates the scalability of ConSGapMiner and shows its ability to mine patterns from high dimensional datasets at low supports.

In this paper, we identify a new task for studying the outlying degree (OD) of high-dimensional data, i.e. finding the subspaces (subsets of features) in which the given points are outliers, which are called their outlying subspaces. Since the state-of-the-art outlier detection techniques fail to handle this new problem, we propose a novel detection algorithm, called High-Dimension Outlying subspace Detection (HighDOD), to detect the outlying subspaces of high-dimensional data efficiently. The intuitive idea of HighDOD is that we measure the OD of the point using the sum of distances between this point and itsknearest neighbors. Two heuristic pruning strategies are proposed to realize fast pruning in the subspace search and an efficient dynamic subspace search method with a sample-based learning process has been implemented. Experimental results show that HighDOD is efficient and outperforms other searching alternatives such as the naive top–down, bottom–up and random search methods, and the existing outlier detection methods cannot fulfill this new task effectively.

Discovery of periodic patterns in time series data has become an active research area with many applications. These patterns can be hierarchical in nature, where a higher-level pattern may consist of repetitions of lower-level patterns. Unfortunately, the presence of noise may prevent these higher-level patterns from being recognized in the sense that two portions (of a data sequence) that support the same (high-level) pattern may have different layouts of occurrences of basic symbols. There may not exist any common representation in terms of raw symbol combinations; and hence such (high-level) pattern may not be expressed by any previous model (defined on raw symbols or symbol combinations) and would not be properly recognized by any existing method. In this paper, we propose a novel model, namely meta-pattern, to capture these high-level patterns. As a more flexible model, the number of potential meta-patterns could be very large. A substantial difficulty lies in how to identify the proper pattern candidates. However, the well-known Apriori property is not able to provide sufficient pruning power. A new property, namely component location property, is identified and used to conduct the candidate generation so that an efficient computation-based mining algorithm can be developed. Last, but not least, we apply our algorithm to some real and synthetic sequences and some interesting patterns are discovered.

In this paper we study methods that combine multiple classification models learned over separate data sets. Numerous studies posit that such approaches provide the means to efficiently scale learning to large data sets, while also boosting the accuracy of individual classifiers. These gains, however, come at the expense of an increased demand for run-time system resources. The final ensemble meta-classifier may consist of a large collection of base classifiers that require increased memory resources while also slowing down classification throughput. Here, we describe an algorithm for pruning (i.e., discarding a subset of the available base classifiers) the ensemble meta-classifier as a means to reduce its size while preserving its accuracy and we present a technique for measuring the trade-off between predictive performance and available run-time system resources. The algorithm is independent of the method used initially when computing the meta-classifier. It is based on decision tree pruning methods and relies on the mapping of an arbitrary ensemble meta-classifier to a decision tree model. Through an extensive empirical study on meta-classifiers computed over two real data sets, we illustrate our pruning algorithm to be a robust and competitive approach to discarding classification models without degrading the overall predictive performance of the smaller ensemble computed over those that remain after pruning.

In this paper, we propose two new algorithms for mining association rules between words in text databases. The characteristics of text databases are quite different from those of retail transaction databases, and existing mining algorithms cannot handle text databases efficiently because of the large number of itemsets (i.e., words) that need to be counted. Two well-known mining algorithms, Apriori algorithm and Direct Hashing and Pruning (DHP) algorithm, are evaluated in the context of mining text databases, and are compared with the new proposed algorithms named Multipass-Apriori (M-Apriori) and Multipass-DHP (M-DHP). It has been shown that the proposed algorithms have better performance for large text databases.